CN109859180A - Merge the image set quality enhancing evaluation method of a variety of measurement criterions - Google Patents
Merge the image set quality enhancing evaluation method of a variety of measurement criterions Download PDFInfo
- Publication number
- CN109859180A CN109859180A CN201910072215.7A CN201910072215A CN109859180A CN 109859180 A CN109859180 A CN 109859180A CN 201910072215 A CN201910072215 A CN 201910072215A CN 109859180 A CN109859180 A CN 109859180A
- Authority
- CN
- China
- Prior art keywords
- image
- image set
- enhancing
- quality
- original image
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Abstract
The invention proposes a kind of image set quality for merging a variety of measurement criterions to enhance evaluation method, it is given a mark respectively using each image of the existing a variety of image quality evaluation criterion to original image set first, then mass fraction average and standard deviation is calculated, secondly mass fractions relative is calculated, it considers further that the mass fractions relative under multiple images quality metric criterion, finally obtains the fusion mass score of original image set.The present invention can determine whether the algorithm for image enhancement to the reinforcing effect of original image set by comparing the size of the fusion mass score of enhancing image set and the fusion mass score of original image set.When acting on original image set there are many algorithm for image enhancement, then by comparing the fusion mass score size for enhancing image set under different images enhancing algorithm, optimal algorithm for image enhancement can be picked out for original image set.Compared to existing averaging method, the image set quality enhancing evaluation method for merging a variety of measurement criterions has higher reliability.
Description
Technical field
The present invention provides a kind of evaluation method for merging a variety of measurement criterions for the enhancing of image set quality, belongs to image increasing
By force with image quality evaluation field.
Background technique
The superiority and inferiority needs of algorithm for image enhancement are judged by image quality evaluation criterion.Under normal conditions, to single image
When quality enhancing is evaluated, gives a mark usually using a certain image quality evaluation criterion to enhanced image, pass through ratio
Compared with score value size, to illustrate the superiority and inferiority of algorithm for image enhancement.Equally, existing averaging method enhances the quality of image set and carries out
When evaluation, given a mark first using existing quality metric criterion to each image in enhanced image set, then
It is averaged, the superiority and inferiority for enhancing algorithm is judged according to the size of average value.However the defect of averaging method is it will be apparent that i.e.
It only only considered this index of average value, when image quality scores certain in image set are very high or very low, average value is very
It is easy fluctuation, and existing averaging method can only judge image set under a certain image quality evaluation criterion, no
Multiple images quality metric criterion can be comprehensively considered.With the arriving of big data era, image data is increasingly huge, usually to scheme
Image set is unit to enhance great amount of images.
Summary of the invention
The object of the present invention is to provide a kind of image set quality of high reliablity to enhance evaluation method, screens for image set
A kind of performance preferably algorithm for image enhancement out.
In order to achieve the above object, the technical solution of the present invention is to provide a kind of image sets for merging a variety of measurement criterions
Quality enhances evaluation method, which comprises the following steps:
Step 1: all images that original image is concentrated are carried out respectively using existing a variety of image quality evaluation criterion
Quality evaluation obtains every original image mass fraction under each image quality evaluation criterion, i-th original image is existed
Mass fraction under jth kind image quality evaluation criterion is defined as αij；
Step 2: mass fraction average value and side of the original image set under each image quality evaluation criterion are calculated
Difference, mass fraction average value of the original image set under jth kind image quality evaluation criterion are Uj, variance Sj, then have:
In formula, n is total number of original image in original image set, and m is the total number of image quality evaluation criterion；
Step 3: calculating mass fractions relative of the original image set under each image quality evaluation criterion, original
Mass fractions relative of the image set under jth kind image quality evaluation criterion is COVj, then have:
Step 4: comprehensively consider the mass fractions relative under m kind image quality evaluation criterion, calculate melting for original image set
Mass fraction EV is closed, then is had:
EV=λ1×COV1+λ2×COV2+…+λm×COVm, in formula, λ1, λ2..., λmFor weight coefficient；
Step 5: it concentrates all original images to carry out quality enhancing original image using a certain algorithm for image enhancement E, obtains
To enhanced image set；
Step 6: using image quality evaluation criterion identical with step 1 respectively to the whole in enhanced image set
Enhance image and carry out quality evaluation, obtains mass fraction of the every enhancing image under each image quality evaluation criterion, it will
Mass fraction of i-th enhancing image under jth kind image quality evaluation criterion is defined as βij；
Step 7: calculating mass fraction average value and variance of the enhancing image set under each quality metric criterion,
Enhancing mass fraction average value of the image set under jth kind image quality evaluation criterion is U 'j, variance is S 'j, then have:
Step 8: enhancing image set mass fractions relative under each quality metric criterion is calculated, image set is enhanced
Mass fractions relative under jth kind image quality evaluation criterion is COVj', then have:
Step 9: comprehensively considering the mass fractions relative under a variety of quality metric criterion, calculates melting for enhancing image set
Close mass fraction EV ':
EV '=λ1×COV′1+λ2×COV′2+…+λm×COV′m；
Step 10: when different algorithm for image enhancement acts on a certain original image set, corresponding fusion is obtained
Mass fraction EV ' judges to enhance algorithm for the optimum image of the original image set according to the size of fusion mass score EV '.
Preferably, in step 4, the weight coefficient λ1, λ2..., λmIt can be determined according to specific practical application request,
Default-weight size is
Preferably, in step 10, judgment rule is as follows: (1) for a certain algorithm for image enhancement, if corresponded
Fusion mass score EV ' be less than original image set fusion mass score EV, then the algorithm for image enhancement is generally to image
Collection quality is enhanced, conversely, then the algorithm for image enhancement generally cuts down the quality of image set；(2) for
Different algorithm for image enhancement, corresponding fusion mass score EV ' is smaller, then illustrates the algorithm for image enhancement to image
The reinforcing effect of collection is better.
The present invention is using above-mentioned technical proposal to enhancing image set (a certain algorithm for image enhancement acts on original image set)
It is handled, obtains the fusion mass score of enhancing image set.By comparing the enhancing fusion mass score of image set and original
The size of the fusion mass score of image set can determine whether that algorithm for image enhancement of the invention imitates the enhancing of original image set
Fruit.When acting on original image set there are many algorithm for image enhancement, then by comparing enhancing figure under different images enhancing algorithm
The fusion mass score size of image set can pick out optimal algorithm for image enhancement for original image set.Compared to existing
Averaging method, the image set quality enhancing evaluation method for merging a variety of measurement criterions have higher reliability.
Detailed description of the invention
Fig. 1 is overall framework figure of the invention；
Fig. 2 is the specific flow chart for being the proposed method of the present invention.
Specific embodiment
Present invention will be further explained below with reference to specific examples.It should be understood that these embodiments are merely to illustrate the present invention
Rather than it limits the scope of the invention.In addition, it should also be understood that, after reading the content taught by the present invention, those skilled in the art
Member can make various changes or modifications the present invention, and such equivalent forms equally fall within the application the appended claims and limited
Range.
It is next by taking underwater picture collection as an example that the present invention is further explained.Select 100 underwater pictures as original image set, choosing
These three classical underwater picture quality metric criterion of UCIQE, UIQM, entropy are taken to be illustrated.UCIQE(Underwater
Color Image Quality Evaluation) it is to be proposed by Yang and Arcot, it is current most widely used underwater figure
Image quality metric criterion；UIQM (Underwater Image Quality Evaluation Metric) is by Karen et al.
The classical underwater picture quality metric criterion of a kind of comparison proposed；Entropy criterion is a kind of picture quality based on Shannon information theory
Measurement criterion.Choose algorithm for image enhancement of the histogram equalization algorithm as this example.
Specific implementation steps are as follows:
Step 1: UCIQE is expressed as Q1, UIQM be expressed as Q2, entropy be expressed as Q3, to complete in original image set
Portion image (I1, I2..., I100) carry out quality evaluation.Obtain every original image IiMass fraction is distinguished under UCIQE criterion
For αi1；Mass fraction is respectively α under UIQM criterioni2；Mass fraction under entropy criterion is αi3, wherein i (i=1,
2 ... 100) it is image label, 100 be amount of images included in original image set.
Step 2: mass fraction average value and variance of the original image set under UCIQE, UIQM, entropy criterion are calculated:
UCIQE criterion:
UIQM criterion:
Entropy criterion:
Step 3: mass fractions relative of the original image set under UCIQE, UIQM, entropy criterion is calculated:
UCIQE criterion:
UIQM criterion:
Entropy criterion:
Step 4: comprehensively consider the mass fractions relative under a variety of quality metric criterion, calculate the fusion of original image set
Mass fraction.Due to the quality metric criterion number m=3 that this example is chosen, so λ1, λ2, λ3Take default valueThat is:
Step 5: it concentrates all images to carry out quality enhancing original image using histogram equalization algorithm, is enhanced
Image set (I ' afterwards1, I '2..., I '100)。
Step 6: using the image quality evaluation criterion of step 1 respectively to all images in enhanced image set
(I′1, I '2..., I '100) quality evaluation is carried out, obtaining mass fraction of the every enhancing image under UCIQE criterion is βi1；?
Mass fraction is respectively β under UIQM criterioni2；Mass fraction under entropy criterion is βi3。
Step 7: mass fraction average value and variance of the enhancing image set under UCIQE, UIQM, entropy criterion are calculated:
UCIQE criterion:
UIQM criterion:
Entropy criterion:
Step 8: the mass fractions relative under three quality metric criterion is calculated separately:
UCIQE criterion:
UIQM criterion:
Entropy criterion:
Step 9: comprehensively considering the mass fractions relative under three quality metric criterion, finally obtains enhancing image set
Fusion mass score:
Step 10: calculating by Matlab program, show that the fusion mass score EV of original image set is 0.3653, passes through
The value of the fusion mass score EV ' of enhancing image set after histogram equalization algorithm enhancing is 0.0110.The value of EV ' is less than EV
Value, i.e. histogram equalization algorithm generally enhances the quality of the underwater picture collection, and the value very little of EV ', explanation
Histogram equalization algorithm is fine to the reinforcing effect of the underwater picture collection.
The present invention can provide appraisement system for the various quality enhancement algorithms applied to image set, find one for image set
The suitable algorithm for image enhancement of kind.It is not difficult to find that when a certain algorithm for image enhancement acts on a certain image set, only as EV '
When value is less than EV value, it can just illustrate that the algorithm for image enhancement has practical value to this image set, and have and be applied to reality
Reliability in the environment of border.
Claims (3)
1. a kind of image set quality enhancing evaluation method for merging a variety of measurement criterions, which comprises the following steps:
Step 1: quality is carried out to all images that original image is concentrated respectively using existing a variety of image quality evaluation criterion
Evaluation, obtains every original image mass fraction under each image quality evaluation criterion, by i-th original image in jth
Mass fraction under kind image quality evaluation criterion is defined as αij；
Step 2: calculating mass fraction average value and variance of the original image set under each image quality evaluation criterion,
Mass fraction average value of the original image set under jth kind image quality evaluation criterion is Uj, variance Sj, then have:
In formula, n is total number of original image in original image set, and m is the total number of image quality evaluation criterion；
Step 3: mass fractions relative of the original image set under each image quality evaluation criterion, original image are calculated
Integrate the mass fractions relative under jth kind image quality evaluation criterion as COVj, then have:
Step 4: comprehensively consider the mass fractions relative under m kind image quality evaluation criterion, calculate the fusion matter of original image set
Score EV is measured, then is had:
EV=λ1×COV1+λ2×COV2+…+λm×COVm, in formula, λ1, λ2..., λmFor weight coefficient；
Step 5: it concentrates all original images to carry out quality enhancing original image using a certain algorithm for image enhancement E, is increased
Image set after strong；
Step 6: the whole in enhanced image set is enhanced respectively using image quality evaluation criterion identical with step 1
Image carries out quality evaluation, mass fraction of the every enhancing image under each image quality evaluation criterion is obtained, by i-th
Enhance mass fraction of the image under jth kind image quality evaluation criterion and is defined as βij；
Step 7: mass fraction average value and variance of the enhancing image set under each quality metric criterion, enhancing are calculated
Mass fraction average value of the image set under jth kind image quality evaluation criterion is U 'j, variance is S 'j, then have:
Step 8: calculating enhancing image set mass fractions relative under each quality metric criterion, enhances image set in jth
Mass fractions relative under kind image quality evaluation criterion is COV 'j, then have:
Step 9: comprehensively considering the mass fractions relative under a variety of quality metric criterion, calculates the fusion matter of enhancing image set
Measure score EV ':
EV '=λ1×COV′1+λ2×COV′2+…+λm×COV′m；
Step 10: when different algorithm for image enhancement acts on a certain original image set, corresponding fusion mass is obtained
Score EV ' judges to enhance algorithm for the optimum image of the original image set according to the size of fusion mass score EV '.
2. a kind of image set quality enhancing evaluation method for merging a variety of measurement criterions as described in claim 1, feature exist
In, in step 4, the weight coefficient λ1, λ2..., λmIt can be determined according to specific practical application request, default-weight is big
It is small to be
3. a kind of image set quality enhancing evaluation method for merging a variety of measurement criterions as described in claim 1, feature exist
In in step 10, judgment rule is as follows: (1) for a certain algorithm for image enhancement, if corresponding fusion mass point
Number EV ' is less than the fusion mass score EV of original image set, then the algorithm for image enhancement generally carries out image set quality
Enhancing, conversely, then the algorithm for image enhancement generally cuts down the quality of image set；(2) different images is increased
Strong algorithms, corresponding fusion mass score EV ' is smaller, then illustrates the algorithm for image enhancement to the reinforcing effect of image set
Better.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910072215.7A CN109859180A (en) | 2019-01-25 | 2019-01-25 | Merge the image set quality enhancing evaluation method of a variety of measurement criterions |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910072215.7A CN109859180A (en) | 2019-01-25 | 2019-01-25 | Merge the image set quality enhancing evaluation method of a variety of measurement criterions |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109859180A true CN109859180A (en) | 2019-06-07 |
Family
ID=66896153
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910072215.7A Pending CN109859180A (en) | 2019-01-25 | 2019-01-25 | Merge the image set quality enhancing evaluation method of a variety of measurement criterions |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109859180A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110415217A (en) * | 2019-07-03 | 2019-11-05 | 东华大学 | Enhance preferred method based on subset guiding and the image set of the coefficient of variation |
| CN112102314A (en) * | 2020-11-02 | 2020-12-18 | 成都考拉悠然科技有限公司 | Computing method for judging quality of face image based on uncertainty |
| CN113506260A (en) * | 2021-07-05 | 2021-10-15 | 北京房江湖科技有限公司 | Face image quality evaluation method and device, electronic equipment and storage medium |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2012149645A1 (en) * | 2011-05-04 | 2012-11-08 | École De Technologie Supérieure | Method and system for increasing robustness of visual quality metrics using spatial shifting |
| CN104008543A (en) * | 2014-05-12 | 2014-08-27 | 河海大学 | Image fusion quality evaluation method |
| CN106296669A (en) * | 2016-08-01 | 2017-01-04 | 微梦创科网络科技（中国）有限公司 | A kind of image quality evaluating method and device |
| CN106780434A (en) * | 2016-11-15 | 2017-05-31 | 天津大学 | Underwater picture visual quality evaluation method |
| CN107146220A (en) * | 2017-04-14 | 2017-09-08 | 西北大学 | A kind of universal non-reference picture quality appraisement method |
| CN107770517A (en) * | 2017-10-24 | 2018-03-06 | 天津大学 | Full reference image quality appraisement method based on image fault type |
| CN108022241A (en) * | 2017-12-26 | 2018-05-11 | 东华大学 | A kind of coherence enhancing quality evaluating method towards underwater picture collection |
| CN108335289A (en) * | 2018-01-18 | 2018-07-27 | 天津大学 | A kind of full image method for evaluating objective quality with reference to fusion |
| CN108550140A (en) * | 2018-03-28 | 2018-09-18 | 东华大学 | A kind of confidence level coherence enhancing quality evaluating method towards image set |
-
2019
- 2019-01-25 CN CN201910072215.7A patent/CN109859180A/en active Pending
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2012149645A1 (en) * | 2011-05-04 | 2012-11-08 | École De Technologie Supérieure | Method and system for increasing robustness of visual quality metrics using spatial shifting |
| CN104008543A (en) * | 2014-05-12 | 2014-08-27 | 河海大学 | Image fusion quality evaluation method |
| CN106296669A (en) * | 2016-08-01 | 2017-01-04 | 微梦创科网络科技（中国）有限公司 | A kind of image quality evaluating method and device |
| CN106780434A (en) * | 2016-11-15 | 2017-05-31 | 天津大学 | Underwater picture visual quality evaluation method |
| CN107146220A (en) * | 2017-04-14 | 2017-09-08 | 西北大学 | A kind of universal non-reference picture quality appraisement method |
| CN107770517A (en) * | 2017-10-24 | 2018-03-06 | 天津大学 | Full reference image quality appraisement method based on image fault type |
| CN108022241A (en) * | 2017-12-26 | 2018-05-11 | 东华大学 | A kind of coherence enhancing quality evaluating method towards underwater picture collection |
| CN108335289A (en) * | 2018-01-18 | 2018-07-27 | 天津大学 | A kind of full image method for evaluating objective quality with reference to fusion |
| CN108550140A (en) * | 2018-03-28 | 2018-09-18 | 东华大学 | A kind of confidence level coherence enhancing quality evaluating method towards image set |
Non-Patent Citations (6)
| Title |
|---|
| CHAU YI LI 等: "An Online Platform for UndeImage Quality Evaluationrwater", 《INTERNATIONAL CONFERENCE ON PATTERN RECOGNITION》 * |
| CHONGYI LI 等: "An Underwater Image Enhancement Benchmark Dataset and Beyond", 《ARXIV》 * |
| 刘浩 等: "面向图像集的置信区间内一致性增强质量评价准则", 《电子与信息学报》 * |
| 孙晓帆 等: "面向水下图像集的一致性增强评价方法", 《中国图象图形学报》 * |
| 杨艳春 等: "图像融合质量评价方法研究综述", 《计算机科学与探索》 * |
| 蒋刚毅 等: "图像质量评价方法研究进展", 《电子信息学报》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110415217A (en) * | 2019-07-03 | 2019-11-05 | 东华大学 | Enhance preferred method based on subset guiding and the image set of the coefficient of variation |
| CN112102314A (en) * | 2020-11-02 | 2020-12-18 | 成都考拉悠然科技有限公司 | Computing method for judging quality of face image based on uncertainty |
| CN113506260A (en) * | 2021-07-05 | 2021-10-15 | 北京房江湖科技有限公司 | Face image quality evaluation method and device, electronic equipment and storage medium |
| CN113506260B (en) * | 2021-07-05 | 2023-08-29 | 贝壳找房(北京)科技有限公司 | Face image quality assessment method and device, electronic equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111062951B (en) | Knowledge distillation method based on semantic segmentation intra-class feature difference | |
| CN109859180A (en) | Merge the image set quality enhancing evaluation method of a variety of measurement criterions | |
| CN106203430A (en) | A kind of significance object detecting method based on foreground focused degree and background priori | |
| CN107844785B (en) | A kind of method for detecting human face based on size estimation | |
| CN110189256A (en) | A kind of Panorama Mosaic method, computer readable storage medium and panorama camera | |
| JP5916134B2 (en) | Object detection method and object detection apparatus using the method | |
| CN104517095B (en) | A kind of number of people dividing method based on depth image | |
| CN105513080B (en) | A kind of infrared image target Salience estimation | |
| CN107886507B (en) | A kind of salient region detecting method based on image background and spatial position | |
| CN111524164B (en) | Target tracking method and device and electronic equipment | |
| CN110110694B (en) | Visual SLAM closed-loop detection method based on target detection | |
| CN107248174A (en) | A kind of method for tracking target based on TLD algorithms | |
| CN104537355A (en) | Remarkable object detecting method utilizing image boundary information and area connectivity | |
| CN101251896B (en) | Object detecting system and method based on multiple classifiers | |
| CN109034001A (en) | A kind of cross-module state saliency detection method based on Deja Vu | |
| CN110443279B (en) | Unmanned aerial vehicle image vehicle detection method based on lightweight neural network | |
| CN111091592A (en) | Image processing method, image processing apparatus, electronic device, and readable storage medium | |
| CN106557765A (en) | Note detection means and note detection method | |
| CN107123130A (en) | Kernel correlation filtering target tracking method based on superpixel and hybrid hash | |
| CN107909079A (en) | One kind collaboration conspicuousness detection method | |
| CN105721955A (en) | Video key frame selecting method | |
| CN110147816A (en) | A kind of acquisition methods of color depth image, equipment, computer storage medium | |
| CN103748613A (en) | Anisotropic gradient regularization for image denoising, compression, and interpolation | |
| CN104113687A (en) | Image processing method and electronic equipment | |
| CN110503049B (en) | Satellite video vehicle number estimation method based on generation countermeasure network |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20190607 |
|
| WD01 | Invention patent application deemed withdrawn after publication |